lz4_flex VS zstd

I run a benchmark based on my LZ4 implementation (de/compressor) between cax11 and cx11 since the instances cost the same. The code doesn't use any SIMD.

By the way, the PR removing bounds checks from extend_from_within_overlapping is up: https://github.com/PSeitz/lz4_flex/pull/141

Zap uses LZ4 as it's primary compression algorithm, although I might add ZSTD as an option too. The LZ4 crate i'm using goes over the compression better than I can in the span of a comment. Check it out here.

I wrote something here how hc could work: https://github.com/PSeitz/lz4_flex/issues/21

Yes, the updated benchmarks are here: https://github.com/pseitz/lz4_flex#results-v080-17-05-2021-safe-decode-and-safe-encode-off

I ported the block format to Rust matching the C implementation in performance and ratio.

https://github.com/pseitz/lz4_flex

Following this change count_same_bytes is unsound - it offsets the pointer by the value of candidate without any bounds checks, which may result in out-of-bounds access.

> Doesn't take large amount of GPU resources

This is an understatement, zstd dictionary compression and decompression are blazingly fast: https://github.com/facebook/zstd/blob/dev/README.md#the-case...

My real-world use case for this was JSON files in a particular schema, and the results were fantastic.

This VFS will read a sqlite file after it has been compressed using [zstd seekable format](https://github.com/facebook/zstd/blob/dev/contrib/seekable_f...). Built to support read-only databases for full-text search. Benchmarks are provided in README.

Of course, you may get different results with another dataset.

gzip (zlib -6) [ratio=32%] [compr=35Mo/s] [dec=407Mo/s]

zstd (zstd -2) [ratio=32%] [compr=356Mo/s] [dec=1067Mo/s]

NB1: The default for zstd is -3, but the table only had -2. The difference is probably small. The range is 1-22 for zstd and 1-9 for gzip.

NB2: The default program for gzip (at least with Debian) is the executable from zlib. With my workflows, libdeflate-gzip iscompatible and noticably faster.

NB3: This benchmark is 2 years old. The latest releases of zstd are much better, see https://github.com/facebook/zstd/releases

For a high compression, according to this benchmark xz can do slightly better, if you're willing to pay a 10× penalty on decompression.

xz -9 [ratio=23%] [compr=2.6Mo/s] [dec=88Mo/s]

zstd -9 [ratio=23%] [compr=2.6Mo/s] [dec=88Mo/s]

Compression, synchronization and backup systems often use rolling hash to implement "content-defined chunking", an effective form of deduplication.

In optimized implementations, Rabin-Karp is likely to be the bottleneck. See for instance https://github.com/facebook/zstd/pull/2483 which replaces a Rabin-Karp variant by a >2x faster Gear-Hashing.

Get the data https://publicdistst.blob.core.windows.net/data/root.tar.zst magnet:?xt=urn:btih:84931cd80409ba6331f2fcfbe64ba64d4381aec5&dn=root.tar.zst How to extract https://github.com/facebook/zstd Linux (debian): `sudo apt install zstd` ``` tar -I 'zstd -d -T0' -xvf root.tar.zst ```

I've done that experiment with zstd before.

https://github.com/facebook/zstd/blob/dev/programs/zstd.1.md...

Not sure about brotli though.

If you want this functionality in zstd itself, check this out: https://github.com/facebook/zstd/pull/2349